Posts from China
15 February 2026 - How to Ensure Kernels Actually Overlapped
14 February 2026 - Distributed-Native FFA
08 February 2026 - Attention Engine for Inference
07 February 2026 - Support Blackwell with FFA_FA4 Backend
04 February 2026 - Support Muon QK-Clip
25 January 2026 - Optimize Sparse Attention in FFA
24 January 2026 - Support Native Group Collective Based on DeepEP
21 January 2026 - Dynamic Attention Solver
22 December 2025 - Flash Attention 2 Math Derivation
17 November 2025 - Support Learnable Attention Sink
19 October 2025 - Long-Context Attention Benchmark
21 April 2025 - MagiAttention